Neural network based classification of non-averaged event-related EEG responses.

Classification of non-averaged task-related EEG responses with different types of classifier, including self-organising feature map and learning vector quantiser, K-mean, back-propagation and a combination of the last two, is reported. EEG data are collected from approximately one second periods prior to movement of the right or left index finger. A cue stimulus indicating which hand to use is employed. Feature vectors are formed by concatenating spatial information from different EEG electrodes and temporal information from different time incidents during the planning of hand movement. Power values of the most reactive frequencies within the extended alpha-band (5-16 Hz) are used as features. The features are derived from an autoregressive model fitted to the EEG signals. The performance of the classifiers and their ability to learn and generalise is tested with 200 arbitrarily selected event-related EEG data from a normal subject. Classification accuracies as high as 85-90% are achieved with the methods described here. A comparison of the classifiers is made.

Prediction of the side of hand movements from single-trial multi-channel EEG data using neural networks.

Thirty channels of EEG data were recorded prior to voluntary right or left hand movements. Event-related desynchronization (ERD) was quantified in the 8-10 Hz and 10-12 Hz bands in single-trial data and used as training input for a neural network comprised of a learning vector quantizer (LVQ). After a training period, the network was able to predict the side of hand movement from single-trial EEG data recorded prior to movement onset.